Neural Program Search: Solving Programming Tasks from Description and Examples

TL;DR

This paper introduces Neural Program Search, a method that generates programs from natural language descriptions and examples by combining deep learning with program synthesis techniques, using a specialized DSL and a guided search algorithm.

Contribution

It proposes a novel neural program search algorithm that integrates a domain-specific language and a Seq2Tree model to improve program generation from descriptions and examples.

Findings

Outperforms sequence-to-sequence models with attention baseline

Uses a semi-synthetic dataset for evaluation

Demonstrates effective program synthesis from natural language and examples

Abstract

We present a Neural Program Search, an algorithm to generate programs from natural language description and a small number of input/output examples. The algorithm combines methods from Deep Learning and Program Synthesis fields by designing rich domain-specific language (DSL) and defining efficient search algorithm guided by a Seq2Tree model on it. To evaluate the quality of the approach we also present a semi-synthetic dataset of descriptions with test examples and corresponding programs. We show that our algorithm significantly outperforms a sequence-to-sequence model with attention baseline.